Meta unveils Muse Wonder AI models

Data center server racks symbolize new AI models rollout and rising tech shares for META

Key Points

01Meta (META) debuts Muse Wonder and Muse Wonder Full models
02Muse Wonder offered via Meta (META) Model API and Muse Code
03Pricing set around $4 input, $20 output per million tokens
04Meta (META) shares gain about 4% intraday after the launch

Meta introduces new Muse Wonder models

On October 12, 2026, Meta launched two new large-scale AI models, Muse Wonder and Muse Wonder Full, expanding its Muse product lineup. The company describes these models as its most capable Muse systems to date, positioning them in the competitive market for advanced generative AI.

Muse Wonder is being distributed broadly through the Meta Model API as well as the Muse Code developer product, targeting both application builders and existing Muse ecosystem users. Muse Wonder Full, the larger variant, is restricted to paying Muse Code subscribers and is not available through the Meta Model API at launch.

Integration with the Muse personal AI agent

Meta stated that Muse Wonder will soon serve as the engine for its Muse personal AI agent. The initial integration is planned for Muse subscribers, tying the new model directly into a consumer-facing assistant product.

This planned rollout is intended to bring the new model’s capabilities into everyday use within the existing Muse subscriber base, while the higher-capacity Muse Wonder Full remains focused on developers using Muse Code.

Model scale, context window, and benchmarks

Meta’s documentation lists a 1,000,000-token context window for the Muse Wonder family, supporting long inputs and extended conversations in both consumer and developer deployments. This context length applies across the newly launched models.

Reported evaluation results place Muse Wonder Full near the top of at least one industry intelligence index, with scores described in the mid-50s against peer systems. However, multiple reports note that part of the benchmarking and comparative performance tables are drawn from Meta’s own disclosures and have not yet been widely reproduced by independent testing.

Pricing structure for Muse Wonder

Meta released per-token API pricing for Muse Wonder at roughly $4 per million input tokens and $20 per million output tokens. The pricing tables also cite a cached-input rate of about $0.20 per million tokens, providing a lower-cost option for repeated prompts.

These rates apply to use of Muse Wonder through the Meta Model API and give developers clear cost parameters for integrating the model into applications. Muse Wonder Full access is framed as a premium capability within the Muse Code subscription, with no separate API pricing disclosed.

Market reaction to the launch

The announcement of Muse Wonder and Muse Wonder Full was followed by a positive move in Meta’s share price during intraday trading. Reports cited an approximate 4% gain in the afternoon session on the day of the launch.

The share price reaction highlights investor attention to Meta’s progress in advanced AI models and its efforts to commercialize these systems across consumer agents and developer platforms.

Key Takeaways

01Meta is using the Muse Wonder family to link advanced AI directly into both its consumer Muse agent and developer-focused Muse Code platform.
02A 1 million token context window and transparent per-token pricing indicate an emphasis on large-scale, production-ready use cases.
03Early benchmark results are favorable but rely partly on Meta’s own data, underscoring the importance of forthcoming independent evaluations.

References